John Vial and Hilary Goh are veteran Unearthed hackathon participants and winners, having taken part in over 10 of our open innovation events over the last four years.
Following Unearthed Perth 2016, Hilary landed a job as Geoscientist Machine Learning at Mine Vision Systems, after she was introduced to Founding Director Paul Lucey. This involved applying robotics hardware and software to create 3D underground mine mapping applications to do away with the current pen and paper method.
At Unearthed Perth 2017, John and Hilary's team SOUNDSCAPE won the people's choice award and following the event, John founded SOUNDelve, a startup exploring sound in mining processing plants.
The duo has recently joined forces with other machine learning enthusiasts and have co-founded Perth Machine Learning Group (PMLG), a community of deep learning coders. This is a group of students, industry professionals, entrepreneurs and curious individuals who enjoy studying machine learning and run beginner and advanced 2 hour sessions every Thursday at CORE Innovation Hub.
We recently caught up with the co-founders to find out more about their adventures in machine learning. Here's what they had to say:
Please introduce yourselves and tell us a little about your background.
John: I'm John Vial, my latest venture, as VP Robotics at Project 412 is about bringing automotive autonomy to the mining industry through collaboration and joint ventures. However, I have previously started a company using machine learning for acoustic understanding, been a machine learning researcher at CSIRO and achieved a PhD in robotics at Australian Centre for Field Robotics at The University of Sydney.
Hilary: I’m Hilary Goh and I’m currently exploring new ways to map the physical and digital world for mining. I’m a founder of SOUNDelve with my husband, John Vial, where we delve deeply into old mining challenges with a new perspective.
I graduated with a BSc in Geology at the University of Wollongong and completed my Honours in Economic Geology at the University of Tasmania in 2008. I started out as an exploration geologist working on drilling rigs and geophysical surveys in remote areas, before moving on to work at mine sites for various commodities around Australia.
Previously I was a Geoscientist Machine Learning with Mine Vision Systems. I made the jump into this role after participating in Unearthed hackathons and startup weekends where I collaborated with software devs, robotics engineers and others.
I like to attend startup weekends and hackathon events all over Australia as I enjoy discovering the entrepreneurial side of science. Now that new technology can solve age-old problems how do we get this into the hands of the people who need it the most?
I’m also a member of the State Emergency Service where we help the community deal with natural disasters, as well as a member of Women in Mining WA (WIMWA) and Women in Technology WA (WITWA). Recently I got together with some friends and founded the Perth Machine Learning Group where we help each other code for machine learning through weekly meetups and discussions.
How did you first become interested in machine learning? Why do you think you have become so passionate about it?
John: I first became interested in Machine Learning and statistics during my PhD. I found that lots of interesting problems could be solved using statistics, and it can be very technically interesting.
I think that machine learning is currently changing the way that businesses operate, and will continue to do so, so it's foolish not to be up to speed on the most recent techniques and approaches.
Hilary: I was aware of machine learning through my husbands work as a roboticist/machine learning researcher. But it was only when I attended a hackathon where my team explored the possibility of applying ML to automatically cross correlate coal seams across many drill holes that I saw how practical and transformative this method could be to the way we currently process and analyse geoscience data.
What inspired you to start the Perth Machine Learning Group? When was the group established?
John: It was the start of 2017 and I wanted to learn more about Deep Learning. I had applied it in a few small projects but the literature was vast. I recalled how excellent it was during my PhD to have a group of peers who were all learning the same or similar things at the same time. I had been to a lot of other technical meetups before, and most of them lacked the deep peer to peer learning that you get from a PhD cohort. Most of the other meetup groups provided general audience machine learning discussion that was watered down for public consumption. I think that those sort of meetups definitely have their place and are great, but I was yearning to build a group of peers that I could learn with, not a group of business people to market to.
Thus I decided to try and create a social group with some of my friends that would be like my PhD cohort. I asked Sean and Sarada (whom I'd met through an Unearthed hackathon in Perth) as well as Hilary, and they were very keen to start exploring the space with me. I then thought that there could be others who wanted to do the same, so we spoke with a local co-working industry hub CORE Innovation Hub and they kindly advertised our group on their website. I also went and delivered a 3 hour workshop on machine learning at Curtin University and from those two channels we got our first members.
We built our meetup around a free online course, fast.ai. The brilliant thing about this course is that it doesn't require any knowledge other than programming, so we could allow any person who wanted to get their hands dirty (but might be scared off by maths) to get started, being productive right away.
Hilary: Around March 2017 we got together to talk about machine learning techniques and the latest developments in the field. Our main motivation was to share our learnings and practical coding tips with everyone else and so the weekly Thursday meetup was created.
More specifically for me, my motivation is that PMLG can solve the following problems for Perth:
- That there is a skills shortage of applied machine learning coders locally;
- That there are lots of industry problems that could be examined with ML but no one to solve them; and
- That current upskilling pathways were somewhat long (>1yr) and theory focused.
I believe that PMLG can be a way to build up a talent pool of practical machine learning coders to solve local industry challenges.
What are PMLG's goals for 2018? What do you hope that the group achieves?
John: I think the most important thing is to build a community from which people can start to learn and de-mystify this topic. We believe that we are providing fertile ground that will allow a lot of machine learning based startups to thrive in Perth, and we hope that this year members from our community can go out and start applying machine learning in their companies.
Hilary: The group is still growing and we are continuing our focus on providing a meetup of learning and sharing ML techniques for practical coders.
You have recently received sponsorship from Western Power and CORE Innovation Hub. How did this come about and what have you been working on?
John: Western Power have generously sponsored our group and our space. A Western Power employee, Grant joined our community and started to see that we were straining the space we had found, so he went and asked Western Power to support us through a partnership with CORE Innovation Hub. CORE is now providing us the use of space and even a few events.
Hilary: Since the beginning, members of the Western Power data science teams have been attending the PMLG meetups. One member, Grant, could see the value his team was gaining from learning new techniques and approached the Western Power Executive seeking to help us secure a permanent space for our weekly meetups.
You have recently launched a one day coding course to promote ML to women. Can you shed some light on how this came about and provide further detail?
John: Our Women in Machine Learning workshop ran on 5 May 2018 at CORE and was a sold out event. The course is an introduction to Machine Learning based heavily on the first part of the fast.ai course.
The reason we are doing this is that we found that our group had a big gender bias, and we believe that it is because of a barrier to entry that exists in setting up and getting started. We were worried that the lack of female role models in our group meant that when women came to our group they might subconsciously think machine learning is not for them.
The women in ML event is aimed at women to specifically get them past the early hurdles, and to show them that it's not too hard, and that there exists a huge peer group of successful women in ML to learn with.
Hilary: Our initiative for the Women in Machine Learning workshop came from several members who commented that there was a lack of women in the group. In addition, feedback from our beginner members indicated that getting started (the logistical setup) in machine learning coding was somewhat confusing and tricky. We hope that our workshop helps to overcome these hurdles.
Are there any other projects on the horizon for Perth Machine Learning Group?
Hilary: Some future projects we would like to achieve are:
- Beginner workshops to help people over the initial hurdles of setting up their machine learning coding environments;
- Reach out to various companies to help them up skill their data science teams; and
- Reach out to universities to help their students gain practical coding experience.
Thank you to Hilary Goh and John Vial for taking the time to share their insights into the value of deep and machine learning.
We wish Perth Machine Learning Group all the best in their future endeavours and look forward to seeing what the motivated group achieves.
Are you interested in deep learning and machine learning? Do you prefer to understand by programming? To express your interest in PMLG, join the meetup group: https://www.meetup.com/Perth-Machine-Learning-Group/
For more information about upcoming Unearthed events, visit: https://portal.unearthed.solutions/